Reasoning in criminal intelligence analysis through an argumentation theory-based framework

This thesis provides an in-depth analysis of criminal intelligence analysts’ analytical reasoning process and offers an argumentation theory-based framework as a means to support that reasoning process in software applications. Researchers have extensively researched specific areas of criminal intelligence analysts’ sensemaking and reasoning processes over the decades. However, the research is fractured across different research studies and those research studies often have high-level descriptions of how criminal intelligence analysts formulate their rationale (argument). This thesis addresses this gap by offering low level descriptions on how the reasoning-formulation process takes place. It is presented as a single framework, with supporting templates, to inform the software implementation process.
Knowledge from nine experienced criminal intelligence analysts from West Midlands Police and Belgium’s Local and Federal Police forces were elicited through a semi-structured interview for study 1 and the Critical Decision Method (CDM), as part of the Cognitive Task Analysis (CTA) approach, was used for study 2 and study 3. The data analysis for study 1 made use of the Qualitative Conventional Content Analysis approach. The data analysis for study 2 made use of a mixed method approach, consisting out of Qualitative Directed Content Analysis and the Emerging Theme Approach. The data analysis for study 3 made use of the Qualitative Directed Content Analysis approach.
The results from the three studies along with the concepts from the existing literature informed the construction of the argumentation theory-based framework. The evaluation study for the framework’s components made use of Paper Prototype Testing as a participatory design method over an electronic medium. The low-fidelity prototype was constructed by turning the frameworks’ components into software widgets that resembled widgets on a software application’s toolbar. Eight experienced criminal intelligence analysts from West Midlands Police and Belgium’s Local and Federal Police forces took part in the evaluation study. Participants had to construct their rationale using the available components as part of a simulated robbery crime scenario, which used real anonymised crime data from West Midlands Police force. The evaluation study made use of a Likert scale questionnaire to capture the participant’s views on how the frameworks’ components aided participants with; understanding what was going on in the analysis, lines-of-enquiry and; the changes in their level of confidence pertaining to their rationale. A non-parametric, one sample z-test was used for reporting the statistical results. The significance is at 5% (α=0.05) against a median of 3 for the z-test, where μ =3 represents neutral. The participants reported a positive experience with the framework’s components and results show that the framework’s components aided them with formulating their rationale and understanding how confident they were during different phases of constructing their rationale.